Created
January 3, 2015 18:42
-
-
Save willtownes/e951d567633815656837 to your computer and use it in GitHub Desktop.
Elastic Net Cp function
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
#sigma2, df, and Cp may not be reliable in default elasticnet function enet | |
#provide replacements | |
#S2L is the estimate for sigma2 based on maximal model (full_ols_model from lm()) | |
S2L<- sum(residuals(full_ols_model)^2)/full_ols_model$df.residual | |
Cp_enet<-function(enet_obj,y,X,S2L){ | |
#returns the Mallows Cp statistics for a given elastic net object | |
betamat<-enet_obj$beta.pure | |
#create a matrix where each column is prediction at one of the df levels | |
yhats<-X%*%t(betamat) + mean(y) | |
resids<-yhats-y | |
ESS<-colSums(resids^2) | |
lambda<-enet_obj$lambda | |
#indices of which columns are in active set | |
actives<-unlist(enet_obj$actions) | |
k<-ncol(X) | |
df<-c(1,rep(NA,k)) | |
#df is determined by trace of hat matrix | |
for(i in 1:k){ | |
Xactive<-X[,actives[1:i]] | |
H = Xactive%*%solve(t(Xactive)%*%Xactive+diag(lambda,i),t(Xactive)) | |
df[i+1] = 1+sum(diag(H)) #compute trace | |
} | |
n<-length(y) | |
Cp<-ESS/S2L-n+2*df | |
return(Cp) #a vector | |
} |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment